Postgres 19 Feature Preview: Sequence Synchronization for Logical Replication

Logical replication in Postgres has a long-standing limitation: while table data replicates perfectly, the sequences backing SERIAL and IDENTITY columns stay at their initial values on the subscriber. This creates a problem during upgrades or failover—promote a subscriber to primary, and the next INSERT tries to generate an ID that already exists, triggering a primary key violation.

Postgres 19 solves this with automatic sequence synchronization, introduced across three commits by Amit Kapila:

96b37849 added publication support for sequences
f0b3573c introduced the REFRESH SEQUENCES command
5509055d implemented the synchronization worker.

Usage#

Postgres 19 adds FOR ALL SEQUENCES to publications:

-- Publisher
CREATE PUBLICATION upgrade_pub FOR ALL TABLES, ALL SEQUENCES;

-- Subscriber
CREATE SUBSCRIPTION upgrade_sub
    CONNECTION 'host=publisher-db dbname=production'
    PUBLICATION upgrade_pub;

A new sequencesync worker automatically synchronizes sequences in batches. Sequences track their state (INIT → READY) in pg_subscription_rel.

SELECT srsubid, srrelid::regclass, srsubstate
FROM pg_subscription_rel
WHERE srsubstate IN ('i', 'r');  -- 'i' = INIT, 'r' = READY

Sequences only sync when you explicitly run REFRESH SEQUENCES. Build REFRESH SEQUENCES into your cutover runbook.

ALTER SUBSCRIPTION upgrade_sub REFRESH SEQUENCES;

Closing Thoughts#

Postgres is systematically closing logical replication gaps. Postgres 18 added generated column support. Now Postgres 19 adds sequence synchronization, the largest hurdle in zero-downtime upgrade and failover workflows.

The remaining major limitation is DDL replication—schema migrations still require manual coordination. But for upgrade and failover scenarios, this is less critical than sequence synchronization. When promoting a subscriber, you typically want identical schemas already in place. Missing sequence values cause immediate production failures; managing DDL changes is standard operational procedure. Let's save this for the future release.